Table 6.

Impact of individual-level uncertainty on lung cancer risk prediction performance

Prediction models	PRS CI-based (n = 1256)	PRS mean-based (n = 6012)
	AUC (95% CI)	AUC (95% CI)
PRS > 90th vs. PRS < 10th (PRS risk)	0.6076 (0.6055, 0.6082)	0.5925 (0.5903, 0.5948)
PRS risk + age	0.6363 (0.63196, 0.63688)	0.6094 (0.6093, 0.6097)
PRS risk + age + gender	0.6347 (0.6198, 0.6407)	0.6088 (0.6041, 0.6106)
PRS risk + age + gender + smoking status	0.7104 (0.6980, 0.7191)	0.6884 (0.6852, 0.6905)
PRS risk + age + gender + packyears	0.7299 (0.7179, 0.7388)	0.7182 (0.7154, 0.7202)
Age + gender + smoking status	0.6584 (0.6562, 0.6610)	0.6602 (0.6558, 0.6621)

The prediction model performances incorporating different risk factors of PRS-based risk subgroup, age, gender, and smoking history were evaluated in subsets of individuals that were identified by the PRS CI-based approach and by the PRS mean-based approach. For the PRS CI-based approach, the models were constructed and evaluated in the individuals that can be identified with certainty (n = 1256). As a comparison, we constructed the same models and evaluated in the individuals that were classified as the lowest and highest risk by PRS-16-CV mean (n = 6012). The model performance was evaluated using five-fold cross-validation. Area under the curve (AUC) and 95% confidence intervals (CI) are shown